.
Does anyone out there know if this is a safe change? I would appreciate any
feedback/advice here. If I don't hear any objections, I'll file a sub-task to
DERBY-805 for this issue and post the above patch there.
Many thanks in advance to anyone who might have input/feedback here.
Army
Army wrote:
This hasn't really been an issue to date because the best plan chosen
by the subquery is typically independent of the outer query's current
permutation--with the exception of outerCost, which is passed in from
the outer query and is factored into the subquery's cost estimates
questions?
Army
Army wrote:
Dyre Tjeldvoll (JIRA) wrote:
[ http://issues.apache.org/jira/browse/DERBY-815?page=all ]
Dyre Tjeldvoll updated DERBY-815:
-
Attachment: derby-815_2.diff
derby-815_2.stat
derbyall_report_2.txt
New patch
for your patience,
Army
Army wrote:
Okay, so maybe that'not as clear as I'd like it to be. However, when I
post these changes to DERBY-805 (as Phase 1 v2), I will also update
the DERBY-805 document to include this description and a walk-through
example showing how it's all supposed to work.
I just attached the new
/reactions/feedback/corrections at all would be greatly appreciated...
Thanks,
Army
to push predicates
containing complex expressions, that's another enhancement of its own. I won't
be doing that with my DERBY-805 changes.
Thanks again--I can't say that enough--for reading the document. It's a huge
one and I'm grateful for your time and feedback.
Army
pointers...
Army
P.S. In case you're interested, I see the following results when using the
classes directory. Note that the changes I hope to make for the 10.2 release
will catch the class not found type of exceptions and print something more
friendly. In the meantime, though:
Sun 1.3
this document and provide feedback, direction, or
suggestions, I would be grateful. As I myself am still trying to learn all the
subtleties of Derby optimization, the more feedback I get, the better...
Thanks,
Army
the scoped predicates for re-use, so that we don't end up re-scoping the
same predicate for the same children for every permutation.
Army
category as
D and T. They're for timestamp, date, and time, respectively. So whatever
happens to D and T should probably (I assume, having not investigated
further) also occur for TS. I.e. if you're going to remove D and T, it
seems (to me) that you should probably remove TS, as well...?
Army
column is expected,
etc. etc. can cause existing applications a lot of problems--see DERBY-107. Our
JDBC tests might not catch it, but my past experience is that ODBC applications
(esp. Microsoft ones like Access) will throw a fit if something is even slightly
off...
Army
Kathey Marsden wrote:
This vote is for establishing Bryan Pendleton as a committer for Derby.
Please vote +1 if you approve of Bryan as a committer.
+1
Army
think it might
be helpful for future developers if the new code contained some additional
explanatory comments...
Army
-491 could be _either_ a protocol exception or a
hang, depending on the situation), I think that'd be useful. Not obligatory,
but useful...
Thanks for your patience with my picking,
Army
for people to see what's
going on. Otherwise, I myself am not quite clear on what is actually happening
with the scenario you describe, so it's hard to offer any suggestions...
Army
feedback...
Army
for
sub-optimal performance, because it disallows hash joins where they could
potentially be useful.
Thanks for your patience with my questions on this topic; I'm just trying to get
a grasp on how this all is supposed to work...
Army
Army wrote:
- after considering my explanation of the bug, do you have an idea of a
better way to write a test for it? If not, should I just remove the
DERBY-125 test from prepStmt.java?
Not yet; I'll have to think about this some more.
I have just posted a starting point for a test
Satheesh Bandaram wrote:
Army wrote:
So if we have a subquery, in which case childResult will (or least can)
be a SelectNode, the fact that SelectNode is NOT an instance of Optimizable
means that this method will return false and the optimizer won't ever
consider a hash join when the inner
understanding. Is there a different area that you
can think of where flattening should be occuring? Can you give more info about
where that flattening needs to be done?
BTW, in your example the subquery contains a Cartesian product. Is that
intentional?
Just an example ;)
Army
?
The reason I ask is because I started playing around with making the optimizer
consider hash joins for subqueries, and I have made good progress in that area.
Before I go too far, though, I'm wondering what it is that I'm missing--i.e.
why was this disabled?
Anyone know?
Army
causing damage.
I will try this out and see what happens. In the meantime, thanks again for
being so thorough with all of this. I'll post again when I have further
feedback to give...
Army
, sounds good.
Army
Army wrote:
Bryan Pendleton updated DERBY-125:
--
Attachment: changes2.html
svn-dec_28_2005.diff
[ snip ]
I think these changes are now fully ready for review. Please let me
know your comments as I am very interested to hear what you
is always an quality node, it is not based upon the type
being pushed.
[ snip code replacement ]
I'm far from an optimizer expert, but this change makes sense to me and appears
to fulfill the copy this predicate notion that this block of code is dealing with.
For what that's worth,
Army
of the excellent work,
Army
Bryan Pendleton wrote:
Hi Bryan, This issues are related to DSS chaining which is something
separate than DERBY-614.
Army made a great Wiki page on the subject.
http://wiki.apache.org/db-derby/DssProtocolErrors
Hi Kathey, thank you very much. I will go read Army's page to learn more
= 1001;
But it'd be best to leave the others off. I'll add my choice of XML ids as part
of my XML patch, and other people who implement other types can add the
appropriate ids with their patches, as well.
Army
Army wrote:
I'll add my choice of XML ids as part of my XML patch, and other people
who implement other types can add the appropriate ids with their
patches, as well.
That is, until DRDA is updated to define the correct values, at which point the
temporary values will all have to change
()?
The API for that method says:
Returns the fully-qualified name of the Java class whose instances are
manufactured if the method ResultSet.getObject is called to retrieve a value
from the column.
So should a call to that method fail, too?
Army
by this
method (emphasis is my own). Then we'd at least be returning a valid JDBC 3.0
class name. When JDBC 4.0 support is available, changing to java.sql.SQLXML
seems okay to me; i.e. there wouldn't be any compatibility problems there, would
there?
Army
versions are substituted in)?
Army
$TestFailedException:
fetch did not time out. Highest execution time: 7000 ms
In other words, the call to SET QUERY_TIMEOUT failed with a syntax error and
was ignored, and thus the statement never timed out.
Does that sound correct?
Army
think.
But all of that is based on my examination of the metadata codepath; if I've
overlooked or oversimplified, people should feel free to correct me.
Hopefully that's more helpful than confusing,
Army
Satheesh Bandaram wrote:
Thanks Army for the good description... But I do think there are SOME metadata
calls that are stored in the database...
Yes, you're absolutely right. Just after I sent that last email, I realized one
other check that I should have done before clicking send.
I tried
Rick Hillegas wrote:
Here are the datatypes I want to add, re-enable, or extend in the
10.2 timeframe:
[ snip ]
SQLXML
How would this (the SQLXML type) differ from the existing XML type available
as of 10.1.1 (DERBY-334)?
Army
the way.
Thanks for the info,
Army
and provide any
feedback, that'd be great.
Thanks,
Army
== '\n'))
break;
}
The if statement assumes a slash indicates the end of the line, but as Mamta
pointed out, that's not true. This logic needs to fixed to do the correct
thing. I'll look into it if no one else already is...
Army
of a
reason, Yes, it would be safer to use standarad Java character reading classes.
I'll make that change and post a patch soon.
Army
://svn.apache.org/repos/asf/db/derby/code/trunk
Is that acceptable? If not, what are my alternatives?
Army
Deepa Remesh wrote:
Hi Army,
I have merged a patch which adds new files (DERBY-518). I did not have
any problems. So I just tried your merge command and it worked for me.
Here is the output I get:
Daniel John Debrunner wrote:
I just tried merging a different change 320762 that add
file reflects the fact
that the comments were NOT stripped. For the 10.1 branch, the comments are
still stripped, hence the diff you're seeing.
I'll post a fix for this against 10.1 shortly.
Army
should work).
Can anyone confirm one way or the other? If I don't hear otherwise, I'll file a
Jira issue tomorrow.
Thanks,
Army
Army wrote:
If I attempt to call PreparedStatement.setNull() and specify
Types.VARCHAR when the actual parameter type (as determined at bind
time) is CLOB, Derby in embedded mode will throw an error, while Derby
in client/server mode will succeed.
[ snip ]
Looks like the problem
directories with quoted names and thus the
behavior would (hopefully) be the same across platforms...
Am I correct in thinking that would resolve this issue?
Army
Kathey Marsden wrote:
There are currently 5 issues outstanding for 10.1.2 for Army, Knut and Jeff
[ snip ]
Questions
Do all the open 10.1.2 bugs look like they'll make it for 10.1.2?
I've unassigned myself from DERBY-500 since I'm not currently working on it and
since Sunitha has
I don't know if the
results are different there.
Hope I'm not being too picky...
Army
David Van Couvering wrote:
Hi, Rick. Once this gets approval from Army and Shreyas I'll go ahead
and check it in.
David
Rick Hillegas wrote:
I have attached a new rev of the fix to bug 171. This rev addresses
Army's concerns.
Cheers,
-Rick
The new patch addresses the concerns I brought
things straight in my
own head...
Hope that's more helpful than it is confusing/frustrating...
Army
for it...
Thanks for any feedback/input,
Army
,
slavic, cs);
Hope that helps,
Army
to
comments/suggestions...
Thanks,
Army
) Should I change this to a documentation error, then?
Thanks,
Army
Daniel John Debrunner wrote:
Army wrote:
1) Does this mean that we don't support _any_ triggers that are defined
on tables having lob columns? Or is there some qualifier to that?
Sorry, I think the qualifier should have been: accessing LOB columns of
the modified table by the action
tried preparing these statements using JDBC, but they failed there with
the same error. I assume this is just a documentation error, but if anyone
knows differently, please speak up...
Thanks,
Army
Ooops, copy-paste error.
ij select a from t1 union select b as a from t2;
I// Now it's A instead of 1.
---
1
2
The I result column name above is actually A, I just copied from the wrong
window. Sorry for any confusion.
Army
--
Army wrote:
I noticed
to
the next without any changes to the app code),
3) standards-compliant (based on SQL/XML standards)
I hope all of that makes sense, but if you have any further questions, please do
ask...
Thanks!
Army
to my own personal opinion; I'm
hoping that if anyone out there disagrees with any of this feedback (including
you, Jeff :), s/he will speak up...
Army
at #4...
Thanks again for your interest in working with Derby, and I look forward to
hearing from you as you work on this project. And of course, if you have
questions, please do ask!
Army
:
jdbc:derby:sample;createFrom=c:\mybackups\sample
I think that's it; thanks for your patience, and sorry again for my
incorrect/incomplete comments the first time around...
Army
for this bug have been
handled now, so I give this a +1 to be committed...
Army
, and that the whole section on Identity column attributes might be
more helpful if it was moved to the place where the actual syntax is described...
*shrug*
Army
The attached patch updates a master/canon for the jdbcapi/parameterMapping.java
test; it looks like it was inadvertently left out of svn revision 190127.
Could a committer please commit?
Thanks much,
Army
Index:
java/testing/org/apache/derbyTesting/functionTests/master/DerbyNet/ver2.6
, though, that a slight dblook modification should be made for DERBY-337;
I'll post a comment to that issue and that can serve as the start of a new thread...
Army
, and thanks for figuring that out...
Army
either way...
Army
they are a bit odd, I'm
still okay with them being there...
Thanks again for the review!
Army
.
Does this new XML feature require a VOTE before commit?
Army
if (returnType == null) ...
But I've added comments to the IF, so I guess either is probably okay...
Any preference?
Thanks for the review,
Army
of the server tests.
Most of the masters were correctly updated, but it looks like this one might
have slipped by...
Army
).
Attached is a small patch that updates these two master files to match the
changes committed with DERBY-305. Could a committer please commit?
Thanks,
Army
Index:
java/testing/org/apache/derbyTesting/functionTests/master/DerbyNet/ibm14/importExport.out
)...
Comments?
Army
Andrew McIntyre wrote:
Hello derby-dev,
Here's the current status as we head towards a 10.1 release.
[ snip ]
Derby-337: dblook should generate SQL statements for SQL functions - any takers?
Unless anyone objects, I'll volunteer to look at this one...
Army
is a very simple patch that modifies the following two files by adding
an if (SanityManager.DEBUG) block around the code in question:
org/apache/derby/iapi/sql/dictionary/ColumnDescriptor.java
org/apache/derby/impl/store/raw/data/FileContainer.java
Could a committer please commit?
Army
Index: java
are not. And according to
the documentation, they're not the same. What this JIRA entry is asking for is
new functionality, as opposed to a fix for broken functionality.
Okay, okay, it's a technicality. But still...
*shrug*
Army
Samuel Andrew McIntyre (JIRA) wrote:
[ http://issues.apache.org/jira/browse/DERBY-194?page=all ]
Samuel Andrew McIntyre resolved DERBY-194:
--
Resolution: Fixed
Patch was committed with svn revision 179839. Could Army or George please
: ConcurrentImplicitCreateSchema jdk1.4.2_07 2005-06-04 07:34:15 ***
No master file was found.
Test Failed.
*** End: ConcurrentImplicitCreateSchema jdk1.4.2_07 2005-06-04 07:34:16 ***
Can you (or some other committer) look at this?
Army
to types.iapi, and 2) is
simplified to just worry about a single XML (UTF-8 based) implementation.
Thanks much for the feedback!
Army
/en-us/odbc/htm/odbctransfer_octet_length.asp
I have run derbyall on a Windows 2000 machine with Sun JDK 1.4.2 and all tests
passed.
Could a committer please review/commit this patch?
Thanks,
Army
implementation can deal with it ;)
Otherwise, I'm certainly open to ideas on how else to get this info without
using the SQL layer...?
Army
without running derbyall is a bit too
risky, since my changes weren't just to the new XML files).
Army
be affected.
Unless someone objects (if you do, say so now!), I vote +1. Could a committer
please commit?
Thanks,
Army
out there for review right now, but since this
patch (namely, the new largeData suite) could have an impact on future
tests/patches in this area, I'm hoping some people can review this and let me
know if I need to change anything...
Feedback is appreciated,
Army
M java\drda\org\apache
...
Army
change the bug description to have the correct size
instead of 64K?
Yep, will do.
Army
be on the assumption that this was removed. If it's intentional,
then can you tell me briefly what this is doing?
Thanks for your patience,
Army
think you can go ahead and do things the way you think is best.
So please feel free to make the change as you prefer, and to post the patch to
the list so we can proceed.
Thanks!
Army
updates in the patch. The svn stat output is attached to this
email, along with the patch itself.
Could someone review/commit?
Many Thanks,
Army
M java\engine\org\apache\derby\impl\jdbc\metadata.properties
M java\engine\org\apache\derby\iapi\types\DataTypeUtilities.java
M java\engine
at DERBY-194 and DERBY-319, which are minor metadata changes.
I have patches for both, but am still in the process of running derbyall.
Again, both are fairly minor and should _not_ be considered show-stoppers.
Army
like that could be done...
In short, I approve of the fix, but I'm not sure the corresponding test is fully
reliable. Is there any way to guarantee the test will catch a failure in this
area in the future?
Army
...?
The safest (consistent) thing to do is to return NULL for columns that are
GENERATED BY DEFAULT, so I think that's what I would vote for (despite my
initial email saying otherwise). But that's just me--I hope anyone who feels
otherwise will post saying so...
Army
Oops. In that last email, replace all occurrences of DefaultNodeImpl with
DefaultInfoImpl. For some reason the wrong file name was in my head.
Army wrote:
That said, I think returning any non-null value in
DefaultNodeImpl.toString() will solve the problem The question then
becomes what
be safer to try to remove those characters from
the patch before submitting, if that's possible...?
I will file a JIRA entry for the DRDA protocol problem I mentioned above. You
may want to wait until that protocol issue can be resolved before proceeding
with this DERBY-308 patch...
Army
Army wrote:
TomohitoNakayama wrote:
I have just done DERBY-308.
I uploaded patch just for reviewing.
[snip]
when I myself
tried to run dblook_test_net against the server, I got a protocol
exception with the GENERATED BY DEFAULT column.
[snip]
I will file a JIRA entry for the protocol
Army wrote:
Please find attached the patch for adding initial XML support to Derby.
[ snip patch description ]
I ran the derbylang suite with Sun JDK 1.4.2 on Windows and all
of the tests passed. I haven't had a chance to run the full
derbyall suite yet, but plan to do that tonight.
I
Army wrote:
I looked at this patch and it seems right, but I'd want to be able to
run your new test before giving my full +1. That said, I'm unable to
run the test because I can't get the patch to apply to my codeline.
Could you perhaps re-generate the patch with the latest codeline
files in derbynetmats.runall will be run as part of derbynetclientmats,
as well. Thus, if you add your new test to derbynetmats.runall, you will NOT
have to add it to derbynetclientmats.runall; it will be included by recursion.
Hope that helps,
Army
201 - 300 of 339 matches
Mail list logo